BEGIN:VCALENDAR
VERSION:2.0
PRODID:Linklings LLC
BEGIN:VTIMEZONE
TZID:Europe/Stockholm
X-LIC-LOCATION:Europe/Stockholm
BEGIN:DAYLIGHT
TZOFFSETFROM:+0100
TZOFFSETTO:+0200
TZNAME:CEST
DTSTART:19700308T020000
RRULE:FREQ=YEARLY;BYMONTH=3;BYDAY=-1SU
END:DAYLIGHT
BEGIN:STANDARD
TZOFFSETFROM:+0200
TZOFFSETTO:+0100
TZNAME:CET
DTSTART:19701101T020000
RRULE:FREQ=YEARLY;BYMONTH=10;BYDAY=-1SU
END:STANDARD
END:VTIMEZONE
BEGIN:VEVENT
DTSTAMP:20260724T151407Z
LOCATION:Bldg. 6 - 103
DTSTART;TZID=Europe/Stockholm:20260701T093000
DTEND;TZID=Europe/Stockholm:20260701T100000
UID:submissions.pasc-conference.org_PASC26_sess153_msa275@linklings.com
SUMMARY:ICON4Py: An ICON Performance Portable Implementation in Python Usi
 ng gt4py
DESCRIPTION:Christoph Müller (MeteoSwiss) and Nicoletta Farabillini (ETH Z
 urich / CSCS)\n\nICON4Py represents a significant shift in the development
  of the ICON weather and climate modeling framework, transitioning from a 
 Fortran codebase to a performance-portable Python implementation. The proj
 ect combines the Python flexibility with the performance required for larg
 e-scale numerical weather prediction and climate modeling. A multi-tiered 
 testing strategy ensures numerical consistency with the original Fortran m
 odel.\n\nICON4Py leverages the GT4Py stencil-based model to implement atmo
 spheric computations in a hardware-agnostic manner by enabling automatic c
 ode generation and low-level parallelization and optimization for diverse 
 backends on both CPUs and GPUs. This approach preserves the algorithmic st
 ructure of the original ICON model while improving code readability, maint
 ainability, interoperability with AI tools such as JAX, and rapid prototyp
 ing capabilities.\n\nThe GT4Py-based dynamical core has shown to meet and 
 exceed the performance of the original ICON OpenACC implementation with up
  to 1.5 speed-up. Initial achievements include successful execution of tes
 t cases (e.g., Jablonowski-Williamson) on single GPUs at 20 km resolution.
 \n\nCurrent results demonstrate that the refactored dynamical core is capa
 ble of supporting global kilometer-scale simulations, establishing a found
 ational framework for next-generation, exascale-ready climate modeling.\n\
 nDomain: Climate, Weather, and Earth Sciences, Computational Methods and A
 pplied Mathematics\n\nSession Chairs: Georgiana Mania (DKRZ) and Xavier La
 pillonne (MeteoSwiss)\n\n
END:VEVENT
END:VCALENDAR